With the scans implemented using dual or multi-energy CT systems, the primary objective is to extract the multi spectra information, which is contained in the image data from the different x-ray energy spectra. Low noise input images from the two spectra are needed herefor, since the pixel-related differences between the images of different spectra are relatively minimal. The reconstructions implemented for this purpose are nowadays generally implemented using very soft convolution kernels. A compromise is already accepted here between resolution and noise. Nevertheless the image noise is generally still too great for a meaningful dual or multi-energy evaluation.
It is known from the prior art to apply edge-containing noise reduction methods to the image data recorded with different spectra in order to process this CT image data. In respect of this method, reference is made by way of example to the publication DE 10 2004 008 979. The disadvantage of the procedure described there is that not all the information contained in the data is used, but any spectrum is instead considered to be an independent scan and is treated in respect of its noise reduction.